Simulating the Interaction of a Renewable Portfolio ... · Renewable Portfolio Standard with...
Transcript of Simulating the Interaction of a Renewable Portfolio ... · Renewable Portfolio Standard with...
M
Mark C. Thurber is Associate Director at the Program onEnergy and Sustainable Development (PESD) at StanfordUniversity. His research focuses on the use of energy andenvironmental market simulations to educate and informpolicy making, the role of state-owned enterprises in fossil
fuel production, and how to deliver energy to low-incomepopulations. Dr. Thurber co-edited and contributed to Oil
and Governance, a 2012 volume on national oilcompanies, and The Global Coal Market, which will be
released this year. Dr. Thurber teaches a class (with co-author Frank Wolak) on ‘‘Energy Markets and Policy’’ in
Stanford’s Graduate School of Business. He holds a Ph.D.from Stanford University in Mechanical Engineering
(Thermosciences) and a B.S.E. from Princeton Universityin Mechanical and Aerospace Engineering with a
certificate from the Woodrow Wilson School of Public andInternational Affairs.
Trevor L. Davis is a doctoral candidate in the Program onEnergy and Sustainable Development and the Department
of Economics at Stanford University. His researchinterests include studying the influence of market designon electricity market outcomes. Before coming to Stanford
he worked in a macroeconomic forecasting section at theFederal Reserve Board of Governors and earned a B.A. inEconomics and Statistics from the University of Chicago
and an M.S. in Statistics from George WashingtonUniversity.
Frank A. Wolak is the Holbrook Working Professor ofCommodity Price Studies in the Economics Department
and the Director of the Program on Energy andSustainable Development at Stanford University. He
received his undergraduate degree from Rice University,
and an S.M. in Applied Mathematics and Ph.D. inEconomics from Harvard University. From April 1998 to
April 2011, he was Chair of the Market SurveillanceCommittee (MSC) of the California Independent System
Operator. Dr. Wolak has worked on the design andregulatory oversight of electricity markets in Europe (in
England and Wales, Germany, Italy, Norway and Sweden,
and Spain); in Australia and Asia (in New Zealand,Australia, Indonesia, Korea, the Philippines, and
Singapore); in Latin America (in Brazil, Chile, Colombia,El Salvador, Honduras, Peru, and Mexico); and the U.S.(in California, New York, Texas, PJM, and New England).From January 2012 to December 2014, Dr. Wolak was a
member of the Emissions Market Advisory Committee(EMAC), which advised the California Air Resources
Board on the design and monitoring of the state’s cap-and-trade market for Greenhouse Gas Emissions allowances.
The authors gratefully acknowledge the active andenthusiastic participation in the energy market simulation
by the students in their class on ‘‘Energy Markets andPolicy’’ in the Stanford University Graduate School of
Business. This work was supported by a grant from thePrecourt Institute for Energy and the TomKat Center forSustainable Energy at Stanford University and also by agrant from the International Policy Implementation Lab at
the Freeman Spogli Institute for International Studies(FSI) at Stanford University.
ay 2015, Vol. 28, Issue 4 1
Simulating the Interaction of aRenewable Portfolio Standardwith Electricity and CarbonMarkets
The authors ran a game-based simulation of an electricitymarket with both an RPS and a cap-and-trade market forgreenhouse gas emissions allowances. High renewableenergy shares reduced and shifted the output of thermalunits and pushed down both electricity and carbon prices.The markets for renewable energy, carbon allowances, andspot and forward electricity interacted in complex waysthat are relevant to the behavior of actual markets.
Mark C. Thurber, Trevor L. Davis and Frank A. Wolak
I. Introduction
California’s electricity market is
subject to a renewable portfolio
standard (RPS) and a cap-and-
trade market for greenhouse gas
(GHG) emissions. These policy
instruments establish markets for
renewable energy certificates
(RECs) and greenhouse gas
emissions allowances,
040-6190/# 2015 Elsevier Inc. All rights reserved.,
respectively. Many other
jurisdictions also deploy
overlapping policy tools in
pursuit of different energy,
environmental, economic, and
political goals.1 As the markets
that result become more complex,
they are less amenable to analysis
with theoretical models and
potentially more vulnerable to
strategic behavior that was not
http://dx.doi.org/10.1016/j.tej.2015.04.007 51
A substantial addedbenefit of this new
approach is the abilityto incorporate
the simulation intoshort-duration
workshops.
52
anticipated by the market
designers.
I n an earlier classroom
simulation, we found a
game-based methodology to be
very effective in highlighting
vulnerabilities of wholesale
electricity markets under a
stringent cap and trade (Thurber
and Wolak, 2013). We have now
extended this earlier work to
incorporate the retail side of the
electricity market, intermittent
renewable energy, an RPS, and
critical peak pricing (CPP)
programs by electricity retailers.
From a practical standpoint,
one of the most significant
challenges of simulating these
complex markets is bringing
participants up to a level at which
they can participate in the market
in a sophisticated manner. This
drove us to create an entirely new
platform for the game so that
we could run it in real time in a
classroom and introduce each
new market element to
participants in turn. A substantial
added benefit of this new
approach is the ability to
incorporate the simulation into
short-duration workshops that
alternate game play with
lectures in order to introduce
policymakers and regulators to
important market concepts.
In the remainder of this article
we describe what we found
running the full simulation in a
classroom setting. Because all
trades in the game are executed
by means of our Web-based
clearinghouse, we can obtain a
complete picture of the carbon
1040-6190/# 2015 Elsevier Inc. All rights reser
allowance and REC markets. We
highlight the following
observations, which show the
potential of the game to deliver
policy insights even while
functioning as an educational tool:
F irst, there were significant
interactions in the game
between the markets for
renewable energy, carbon
allowances, spot and forward
electricity, and retail electricity.
A number of these interactions
have been noted in real markets.
However, one advantage of our
simulation game is that features
of the market can be stress-tested
in ways that would be impossible
in the real world. Heavy
penetration of renewable energy
in the game—prompted in part
by the prospect of high REC
prices—effectively shifted and
reduced the output of thermal
units, which produce GHG
emissions when they operate.
This outcome mimicked the so-
called ‘‘duck curve’’ load shape
(see California ISO, 2013) and
crashed electricity prices during
the hours when wind and solar
ved., http://dx.doi.org/10.1016/j.tej.2015.04.007
energy generation was high. High
renewable energy shares also
discouraged retailers from
signing forward contracts for
electricity by creating the
expectation of low spot prices and
blunting the ability of generators
to exercise market power. The
game results highlighted some
important additional nuances. For
example, while high renewable
energy shares tended to push
down the price of both carbon
allowances and electricity prices,
this effect could be totally undone
by the presence of sufficient
market power in the carbon
market.
Second, it was difficult to avoid
some exercise of market power in
both carbon and REC markets.
This was partly due to the limited
number of players in the market:
seven generators and seven
retailers. But there were other
causes too. Initial distributions of
RECs or carbon allowances could
be lopsided, making the exercise
of unilateral market power more
likely. (Our games hinted that
allocating allowances more
evenly among participants might
provide not only a direct dilution
of initial market concentration but
also a ‘‘nudge’’ to trading these
instruments.) And critically, the
number of players either in need
of or able to supply RECs or
allowances naturally tended to
dwindle as the compliance
deadline drew near and teams
completed trades deemed to
be essential to meeting their
compliance obligations. Those
few remaining usually found
The Electricity Journal
It was difficult in ourgame to predict howsupply and demandfor RECs woulddevelop, makingREC prices quitevolatile.
M
themselves with substantial
market power on either the selling
or buying side.
T hird, it was difficult in our
game to predict how supply
and demand for RECs would
develop, making REC prices quite
volatile. There was uncertainty
around how many wind and solar
facilities would be built, how
much energy they would
generate, and how this would
compare to realized electricity
demand. Moreover, the market
power problems mentioned
above often kept REC prices
uncertain even after overall RPS
compliance had become a
foregone conclusion. Policies to
limit price uncertainty—for
example by establishing floor
and ceiling prices for
RECs—would have made RECs a
more predictable source of
revenue for renewable energy
projects in our game. In the real
world, cap-and-trade systems
have benefited from recent
focus on the risks of price
extremes and market power
(Borenstein et al., 2014), while
REC markets—despite suffering
from the same (or worse)
problems—have not yet received
comparable attention.
II. Description of theEnergy andEnvironmental MarketGame
Our original classroom game,
as described in Thurber and
Wolak (2013), added fixed-price
ay 2015, Vol. 28, Issue 4 1
forward contracts and tradable
carbon allowances to the
Electricity Strategy Game (ESG), a
wholesale electricity market
simulation developed by
Borenstein and Bushnell (2011).
The new simulation game is
played on a fully integrated Web
platform that incorporates many
additional market elements.2 We
added retail players that can
directly negotiate the price and
quantity transacted in fixed-price
forward contracts for energy with
generating companies as well as
decide whether to call ‘‘critical
peak pricing rebate’’ (CPP-R)
periods, which reduce the
demand for wholesale electricity
from retail customers. Wholesale
electricity suppliers can now
acquire intermittent wind and
solar units, which when they run
produce RECs that can be sold to
retailers to comply with the RPS.3
In this article we highlight
results from two pairs of games
(2A/2B and 3A/3B4) that
incorporated the full set of
available markets elements.5 Each
pair of games (for example, 2A/2B)
040-6190/# 2015 Elsevier Inc. All rights reserved.,
was played simultaneously, with
each team playing the role of
generating company (‘‘genco’’) in
one game and retailer in the other.
Seven gencos and seven retailers
played in each game. Each genco
held one of seven portfolios of five
to eight dispatchable generating
units each.6 Three gencos were
located in the North zone and four
gencos in the South zone, and there
was a 750 MW transmission line
between zones. When the
distribution of supply and
demand in the North and South
caused the transmission line to
reach its capacity, the North and
South markets separated.
There were also three retail
companies in the North and four
in the South. Unless they signed
fixed-price forward contracts
with gencos, the retail companies
had to procure power from the
short-term market to meet cus-
tomer demand in each hour; they
in turn sold electricity to their
customers at a fixed retail rate of
$100/MWh. A retailer could de-
clare a CPP Rebate in any hour.
The effect of doing this was to
shift the retailer’s demand curve
in by an average of 20 percent,
with some random variation in
the actual effectiveness of the
CPP-R. The retailer then had to
pay $100 for each MWh that
demand was reduced relative to
the retailer’s ‘‘business as usual’’
demand forecast, representing
rebate payments made to
consumers to encourage
conservation. The retailer was still
paid its usual $100/MWh for the
realized demand.
http://dx.doi.org/10.1016/j.tej.2015.04.007 53
Table 1: Intermittent Renewable Energy Technologies in the Gamea
Name Type Location Capacity
(MW)
Overall Capacity Factor
(Wind/Sun as Forecast)Hourly Capacity Factor (Wind/
Sun as Forecast)
Variable Cost
($/MWh)
Fixed
Cost ($/hr)
4am 10am 4pm 10pm
Altamont Pass Wind North 800 25% 40% 10% 10% 40% $0 $10,000
Shiloh Wind North 800 25% 40% 10% 10% 40% $0 $10,000
Tehachapi Wind South 800 25% 40% 10% 10% 40% $0 $10,000
San Gorgonio Wind South 800 25% 40% 10% 10% 40% $0 $10,000
Central Valley Solar PV North 1,300 15% 0% 30% 30% 0% $0 $20,000
Imperial Valley Solar PV South 1,300 15% 0% 30% 30% 0% $0 $20,000
Central Coast Solar PV South 1,300 15% 0% 30% 30% 0% $0 $20,000
Mojave Desert Solar CSP South 1,000 20% 0% 30% 50% 0% $0 $30,000a To represent intermittency in a stylized way, the actual capacity factor in each hour was modeled as a random variable with a truncated normal distribution (no capacity factors less than 0
percent or greater than 100 percent) with standard deviation of 20 percentage points. This resulted in an expected average generation per hour (accounting for the truncation of the
distribution) of 216 MWh for wind, 199 MWh for solar PV, and 202 MWh for solar CSP. We set there to be 100 percent correlation between[178_TD$DIFF] realized capacity factors[179_TD$DIFF] of units of the same type
in [180_TD$DIFF]the [181_TD$DIFF]same [182_TD$DIFF]location [183_TD$DIFF]in the [184_TD$DIFF]same period (e.g., any wind [185_TD$DIFF]unit in the North [186_TD$DIFF]would [187_TD$DIFF]produce [188_TD$DIFF]exactly the same amount of energy in [189_TD$DIFF]a [190_TD$DIFF]given [191_TD$DIFF]hour [192_TD$DIFF]as [193_TD$DIFF]any [194_TD$DIFF]other [195_TD$DIFF]wind unit in the [196_TD$DIFF]North). All other units
were uncorrelated with each other.Any upfront costs of acquiring renewables were, in effect, levelized in the game into an hourly fixed cost. The fixed costs in Table 1 translate into the
following levelized costs of electricity if capacity factors are as forecast: $46/MWh for wind ($10,000/hr divided by 216 MWh/hr), $100/MWh for solar PV ($20,000/hr divided by 199 MWh/
hr), and $149/MWh for solar CSP ($30,000/hr divided by 202 MWh/hr). These levelized costs were chosen to be on the low end of the ranges assembled by Borenstein (2011) for these
technologies, although with[2_TD$DIFF] fixed costs[197_TD$DIFF] for CSP were set relatively higher than[198_TD$DIFF] those for PV due to the limited experience base with [199_TD$DIFF]CSP projects. The prospect for storage is part of the
appeal of real solar CSP developments (Denholm et al., 2013), but we have not yet incorporated the possibility of storage into this game.
54
Games 2A/2B were played in
class and each consisted of two
electricity market ‘‘days,’’ where
each day consisted of four
‘‘hours,’’ each with a different
forecast for both electricity
demand and wind and solar
conditions. Games 3A/3B had
four electricity market days and
were played outside of class over
about two weeks. Before each
electricity market day, gencos
submitted offer prices (indicating
their willingness to supply
energy) for each of their
generating units in each hour of
the game. The intersection of the
resulting aggregate supply
curve with the realized demand
curve (which is downward
sloping but relatively inelastic)
in each hour determines the
short-term price of electricity
1040-6190/# 2015 Elsevier Inc. All rights reser
and quantity of generation in that
hour.
At any point in the game,7
gencos were allowed to acquire an
unlimited number of wind, solar
photovoltaic (PV), and/or con-
centrating solar power (CSP)
units, which have the stylized
characteristics shown in Table 1.
In all hours in the electricity
market after a renewable unit was
acquired, it generated electricity
(and produced RECs) for its
owner at zero variable cost in
accordance with the realized
wind and solar conditions. It also
incurred a fixed cost in each hour
regardless of whether it was
running, consistent with the
actual cost structure of these
generation units.
T he strengths and
weaknesses of each type of
ved., http://dx.doi.org/10.1016/j.tej.2015.04.007
unit were set to be broadly
reflective of how these
technologies might perform in
California. Wind units have the
lowest fixed costs and the highest
capacity factor, and their
generation is typically highest at
night (10 pm and 4 am hours).
Solar PV units have higher fixed
costs and generate power during
the day (10 am and 4 pm). CSP
units are most expensive but can
deliver significant power in late
afternoon, when demand in our
simulation is generally highest
relative to supply.
E ach of the games
incorporated an RPS
requiring that retailers procure 20
percent of the total electricity they
sold from wind and solar units or
face a $365/MWh penalty for any
portion of this obligation not
The Electricity Journal
[(Figure_1)TD$FIG]
Figure 1: Benchmark Simulation to Estimate the Required Carbon Price to Meet the CO2
Cap with Different Levels of Wind and Solar Generation. Each [11_TD$DIFF]curve is [12_TD$DIFF]generated by: (1)Adding the [13_TD$DIFF]specified [14_TD$DIFF]number of [15_TD$DIFF]renewable units into the [16_TD$DIFF]system, [17_TD$DIFF]spread [18_TD$DIFF]evenly [19_TD$DIFF]among the[20_TD$DIFF]gencos, (2) Bidding in all [21_TD$DIFF]dispatchable [22_TD$DIFF]units ([23_TD$DIFF]natural [24_TD$DIFF]gas, [25_TD$DIFF]coal, [26_TD$DIFF]hydro, and [27_TD$DIFF]nuclear) at their[28_TD$DIFF]marginal [29_TD$DIFF]costs, [30_TD$DIFF]including the [31_TD$DIFF]variable [32_TD$DIFF]cost of [33_TD$DIFF]carbon at the [13_TD$DIFF]specified [33_TD$DIFF]carbon [34_TD$DIFF]price, (3)Running the [35_TD$DIFF]market with [36_TD$DIFF]electricity [37_TD$DIFF]demand and [38_TD$DIFF]wind and [39_TD$DIFF]solar [40_TD$DIFF]generation [41_TD$DIFF]set to be[42_TD$DIFF]exactly as [43_TD$DIFF]forecast, and (4) Comparing the [44_TD$DIFF]resulting CO2 [45_TD$DIFF]emissions to the ‘‘Business asUsual’’ ([46_TD$DIFF]BAU) [47_TD$DIFF]case [48_TD$DIFF]where [49_TD$DIFF]there is [50_TD$DIFF]no [33_TD$DIFF]carbon [34_TD$DIFF]price and [50_TD$DIFF]no [38_TD$DIFF]wind and [39_TD$DIFF]solar [51_TD$DIFF]capacity. As [52_TD$DIFF]more[15_TD$DIFF]renewable [22_TD$DIFF]units are [53_TD$DIFF]added to the [16_TD$DIFF]system, a [54_TD$DIFF]lower [33_TD$DIFF]carbon [34_TD$DIFF]price is [55_TD$DIFF]required to [56_TD$DIFF]meet a [57_TD$DIFF]given[45_TD$DIFF]emissions [58_TD$DIFF]reduction [59_TD$DIFF]target.
M
covered by renewable energy
certificates (RECs).8 Generating
companies needed to cover their
CO2 emissions over the entire
period of the game with sufficient
carbon allowances or face a
penalty of $500 for each ton of
CO2 emitted in excess of their
allowance holdings. The total
number of carbon allowances in
the game (the ‘‘cap’’) was set at a
level approximately 30 percent
lower than ‘‘business as usual’’
emissions in the absence of any
carbon regulation.9
The relative stringency of the
RPS and carbon cap is an
important baseline determinant
of REC and allowance prices. The
benchmark simulation (assuming
price-taking behavior by all
ay 2015, Vol. 28, Issue 4 1
market participants) shown in
Figure 1 illustrates the
relationship between the RPS and
the carbon cap in our game. As
has been observed in Europe,
strong renewable energy
mandates push down the carbon
price needed to meet a particular
CO2 cap—potentially all the way
to zero if the RPS itself reduces
emissions to the level of the cap or
below. To create a more
interesting game, we intentionally
set the carbon cap in this market
such that it would not likely be
met by complying with RPS
targets alone. Figure 1 suggests
that the market in our game could
meet the RPS with 8 wind plus 9
solar PV units (representing about
20.5 percent of total generation)
040-6190/# 2015 Elsevier Inc. All rights reserved.,
and then satisfy the cap with an
allowance price of around $70/
ton of CO2.
I n games 2A/2B, all available
carbon allowances were freely
distributed in equal shares to the
gencos at the start of the game. In
games 3A/3B, all of the allow-
ances were auctioned off, with
both gencos and retailers eligible
to bid. RECs and carbon allow-
ances were freely tradable among
teams at any point during the
game,10 including during a ‘‘true
up’’ period after the final day of
the electricity markets and before
any compliance penalties were
assessed.
IIII. Observations fromthe Game
A. Summary of game results
Table 2 illustrates some of the
results in games 2A/2B and 3A/
3B. Renewable energy generation
comfortably exceeded the 20
percent RPS target in all four
games; only in game 2B was it even
close. As would be expected, lower
REC prices were observed in
games that exceeded the RPS tar-
get by a greater margin. Aggregate
emissions came in under the CO2
cap in all games except 2B.
However, the average traded
carbon price remained high in
games 3A and 3B due to significant
market power issues (see Section
III.D). Market power also played a
role in REC markets, but to a lesser
degree, for reasons that we will
discuss.
http://dx.doi.org/10.1016/j.tej.2015.04.007 55
Table 2: Summary of Games [200_TD$DIFF].
Game 2A 2B 3A 3B
Renewable Generation as % of Total 28.3% 22.8% 28.1% 44.1%
Average Traded REC Price $116 $161 $97 $22
CO2 Emissions as % of Cap 92.3% 101.2% 88.8% 74.0%
>50% of CO2 Allowances Initially Held by One Player U U
Average Traded CO2 Allowance Price $112 $298 $290 $304
Average Electricity Price (Generation- [3_TD$DIFF]Weighted) $38 $59 $108 $66[(Figure_2)TD$FIG]
Figure 2: Total Demand Over Each Day in Each Region (Solid Line) and Remaining Demand after Subtracting Renewable Generation in theRegion (Dotted Line), for Game 3B. The [60_TD$DIFF]dark [61_TD$DIFF]gray [62_TD$DIFF]shaded [63_TD$DIFF]areas [64_TD$DIFF]represent [38_TD$DIFF]wind [40_TD$DIFF]generation, and the [65_TD$DIFF]light [61_TD$DIFF]gray [62_TD$DIFF]shaded [63_TD$DIFF]areas [64_TD$DIFF]represent [39_TD$DIFF]solar[40_TD$DIFF]generation. Results [66_TD$DIFF]between the [67_TD$DIFF]simulated [68_TD$DIFF]hours (4 am, 10 am, 4 pm, and 10 pm) are [69_TD$DIFF]interpolated. Wind and [39_TD$DIFF]solar [70_TD$DIFF]represented 44 [71_TD$DIFF]percentof [72_TD$DIFF]total [40_TD$DIFF]generation in this [73_TD$DIFF]game.
56
S pot electricity prices, which
strongly influenced genco
profitability, were a complicated
function of renewable energy
share and carbon price
expectations. As we will discuss
in Section III.B, high wind and
solar penetration tended to push
down electricity prices by
displacing higher-variable-cost
1040-6190/# 2015 Elsevier Inc. All rights reser
units. Expectations of a higher
carbon price would push up
electricity prices as gencos built
carbon costs into their bids.
Such expectations could derive
either from supply-demand
fundamentals in the carbon
market, as in game 2B, or from
concerns that one or two teams
might hold all of the available
ved., http://dx.doi.org/10.1016/j.tej.2015.04.007
allowances, as in games 3A and
3B.
B. Interactions between
renewable energy, carbon, and
electricity markets
As shown in Figure 2 for a
game (3B) in which 44 percent of
total generation came from wind
The Electricity Journal
[(Figure_3)TD$FIG]
Figure 3: Example of the ‘‘Duck Curve,’’ Taken from California ISO (2013), ‘‘Fast Facts: What the Duck Curve Tells us About Managing aGreen Grid’’ (http://www.caiso.com/Documents/Flexibleresourceshelprenewables_FastFacts.pdf). The [74_TD$DIFF]idea is [75_TD$DIFF]that a [76_TD$DIFF]steady [77_TD$DIFF]increase in[39_TD$DIFF]solar [40_TD$DIFF]generation [78_TD$DIFF]over [79_TD$DIFF]time [80_TD$DIFF]takes a [81_TD$DIFF]big [82_TD$DIFF]chunk [83_TD$DIFF]out of [84_TD$DIFF]effective [85_TD$DIFF]daytime [86_TD$DIFF]load, [87_TD$DIFF]creating a [88_TD$DIFF]profile [75_TD$DIFF]that [89_TD$DIFF]looks [90_TD$DIFF]like a [91_TD$DIFF]duck, [92_TD$DIFF]with a [93_TD$DIFF]rapid [94_TD$DIFF]swing in[21_TD$DIFF]dispatchable [95_TD$DIFF]power [96_TD$DIFF]needs as the [97_TD$DIFF]sun [98_TD$DIFF]sets.
M
and solar, renewable generation
could significantly depress the
effective load to dispatchable
units while shifting the time of the
peak. The intermittent nature of
wind and solar also led to
significant hour-to-hour
variability in the load profile. In
real electricity markets like
California’s, the load profile
that results from significant
penetration of solar energy is
known as the ‘‘duck curve’’
(Figure 3). Our results in Figure 2
do not exactly replicate the duck
curve due to the limited number
of hours we considered, the styl-
ized characteristics of our wind
and solar units, and the fact that
players in our game wisely di-
versified between wind and solar
rather than skewing too heavily
toward solar alone. Wind was
ay 2015, Vol. 28, Issue 4 1
cheaper—and as a result turned
out to be more profitable in most
cases—while acquiring solar was
a bet in part that the 4 pm hour
would continue to yield higher
electricity prices than other
periods. The results show the
value of the simulation for
exploring the kinds of load curves
(whether shaped like ducks or
other animals) that may be
generated by different policy
incentives.
A s in some real markets,
zero-variable-cost
generation from renewables
exerted significant downward
pressure on wholesale electricity
prices (and genco profitability).
Figure 4 illustrates this effect,
comparing the 10 am hour in the
North region in a game with no
renewable energy to the 10 am
040-6190/# 2015 Elsevier Inc. All rights reserved.,
hour in a game with significant
renewable energy and broadly
similar bidding from dispatchable
units (Game 3B, Day 4). In the
latter case, wind and solar
generation alone was sufficient to
meet demand, resulting in an
electricity price of zero.
I n regions and periods with
heavy renewable penetration,
generators stopped trying to
exercise unilateral market power
by submitting offer prices in
excess of the marginal cost of their
thermal units, realizing that it was
often futile because of all of the
zero-variable-cost renewable
generation. Figure 5 shows how
renewable energy discouraged
such attempts. In Day 2 of game
3B (top of Figure 5), the ‘‘Big Gas’’
and ‘‘Old Timers’’ teams in the
South region both bid in all of
http://dx.doi.org/10.1016/j.tej.2015.04.007 57
[(Figure_4)TD$FIG]
Figure 4: Comparison of the 10 am hour in the North Region for a Game with No Wind andSolar Units (Top) and One with a Significant Number of Wind and Solar Units. Wind and[39_TD$DIFF]solar [40_TD$DIFF]generation are [99_TD$DIFF]indicated with a [61_TD$DIFF]gray [100_TD$DIFF]bar (bottom). In [101_TD$DIFF]each [47_TD$DIFF]case, the [102_TD$DIFF]series of [103_TD$DIFF]steps isthe [104_TD$DIFF]bid [11_TD$DIFF]curve, the [105_TD$DIFF]near-[106_TD$DIFF]vertical [107_TD$DIFF]line is the [37_TD$DIFF]demand [11_TD$DIFF]curve, and the [108_TD$DIFF]dotted [109_TD$DIFF]horizontal [107_TD$DIFF]line isthe [35_TD$DIFF]market- [110_TD$DIFF]clearing [34_TD$DIFF]price of [111_TD$DIFF]electricity.
58
their units at $500/MWh in an
effort to push the price to its cap
while still keeping enough
capacity on-line to reap
handsome profits. Without
renewables, or even with lesser
renewable capacity, this strategy
would have been guaranteed to
work due to the high South
demand in the 4 pm and 10 pm
hours. However, in this game it
proved a complete failure at 4 pm
due to heavy renewable
generation and an almost
complete failure at 10 pm, when
these two teams did manage to
spike the price but in so doing
took almost all of their own
capacity off-line (bottom of
Figure 5). There were no
comparable attempts to push the
electricity price to its cap over
the remaining three days of
1040-6190/# 2015 Elsevier Inc. All rights reser
game 3B. (Even so, we have to be
careful about extrapolating a
benefit to consumers; in the real
world, unlike in the game,
generators may have advance
warning through weather
forecasts that renewable
generation will be low in certain
periods, allowing them to bid
very high in these periods.)
L acking the same exposure to
the exercise of market power
by generators, retailers also
stopped hedging their spot price
risk through the purchase of
fixed-price forward contracts
from gencos. (As discussed in
Thurber and Wolak (2013), high
forward contract obligations for
gencos have the beneficial effect
of discouraging gencos from
attempting to exercise market
power because they are unlikely
ved., http://dx.doi.org/10.1016/j.tej.2015.04.007
to benefit and may even be forced
to buy high-priced spot power.) In
a benchmark game we played
without renewable energy (or a
cap-and-trade for carbon),
retailers bought almost 40 percent
of their electricity via forward
contracts with gencos in order to
limit their exposure to genco
market power and potentially
high spot prices. In game 3B, with
its proliferation of wind and solar
units, retailers bought 100 percent
of their power on the spot market.
R enewable generation would
be expected to lower
wholesale electricity prices both
directly, by pushing higher-cost
units out of the generation mix,
and indirectly, by increasing the
likelihood that the carbon cap will
be met and thereby decreasing the
expected carbon cost that
generators factor into their bids.
These basic effects were observed
in our game, but so were more
surprising interactions between
the renewable energy, carbon,
and electricity markets. For
example, the very high renewable
energy share in game 3B largely
held down electricity prices in the
first two days of the game. As
shown in Figure 5, it became clear
to generators that they could not
as easily push up short-term
prices by submitting high offer
prices for their thermal units.
Moreover, renewable generation
in game 3B was sufficiently high
that the carbon cap would
easily be met in the aggregate,
suggesting that the cost of buying
allowances to cover emissions
would be low. These calculations
The Electricity Journal
[(Figure_5)TD$FIG]
Figure 5: High and Unpredictable Wind and Solar Generation (Gray Bars at Left) Makes itMore Difficult for Generators to Successfully Exercise Market Power by Bidding at the PriceCap. ‘‘Big Gas’’ and ‘‘Old Timers’’ [112_TD$DIFF]succeeded in [113_TD$DIFF]spiking the [34_TD$DIFF]price in the 10 pm [114_TD$DIFF]hour, [115_TD$DIFF]but[92_TD$DIFF]with [116_TD$DIFF]little [117_TD$DIFF]benefit to [118_TD$DIFF]themselves, as the [119_TD$DIFF]vast [120_TD$DIFF]majority of their [121_TD$DIFF]thermal [22_TD$DIFF]units [122_TD$DIFF]did [123_TD$DIFF]not [124_TD$DIFF]run ([104_TD$DIFF]bid[125_TD$DIFF]curves [126_TD$DIFF]shown for [127_TD$DIFF]game 3B, [128_TD$DIFF]day 1, [129_TD$DIFF]south region) [130_TD$DIFF].
M
initially led gencos to bid in the
zero-carbon-price marginal costs
of all of their thermal units in Day
2 (see the top bid curve in
Figure 6), resulting in low elec-
tricity prices. However, between
Day 2 and Day 4, all of the gencos
realized that a single team (Big
Coal) was holding nearly all of the
available carbon allowances. This
meant that Big Coal would be able
to extract significant profits in the
carbon market even with CO2
emissions coming in significantly
below the cap. By Day 4,
therefore, gencos began factoring
carbon prices of $200–300/ton
into their electricity market bids,
knowing that they might have to
pay this much or more for
allowances to cover emissions
from their thermal units (see the
bottom bid curve in Figure 6).
Because thermal units still set the
market-clearing price in most
hours, the result was high
electricity prices in the final
ay 2015, Vol. 28, Issue 4 1
market day despite the significant
renewable capacity.
C. Why did gencos acquire so
many renewable units?
The game results raise an
important puzzle: Why did
generators collectively sabotage
themselves by buying so many
renewables?11 Large amounts of
renewable energy significantly
harmed their profitability by
pushing down spot electricity
prices and also by limiting their
ability to exercise market power.
Gencos did benefit from selling
RECs to retailers, but the over-
supply of renewables pushed
down REC values—as well as the
market-clearing electricity prices
that each renewable unit would
be paid. Gencos did not face any
RPS compliance obligations
themselves, and there should in
theory have been little incentive
for gencos to buy renewables to
push down the price of carbon,
040-6190/# 2015 Elsevier Inc. All rights reserved.,
as higher carbon prices actually
benefited most gencos as long
as they incorporated this cost
into their bids. (Thurber and
Wolak (2013) showed that a
higher carbon price actually
enhances profitability of all but
one of the gencos in the game
by pushing up electricity prices
faster than it increases costs
and pushes units out of the
merit order, although it is possible
that not all players appreciated
this fact.)
S everal factors were likely
involved in the extraordinary
renewable energy build-outs that
we saw in almost every run of the
game. First, unlike in the real
California, there is no
transmission interconnection
queue or other regulatory or
investment constraint on how
rapidly new renewable units can
be brought online.12 California
generators may be grateful that
they do not live in a state with
more efficient processes for
interconnecting new wind and
solar facilities!
Second, the RPS in our
game—unlike the real RPS in
California and most other
jurisdictions—had teeth, in the
form of high penalties for
retailers that failed to cover
their 20 percent RPS requirement.
Gencos therefore had
confidence that they could sell
RECs to retailers at good prices,
at least if the RPS target was
not exceeded by too much in the
aggregate. In fact, some teams
intentionally bought up
significant renewables early in
http://dx.doi.org/10.1016/j.tej.2015.04.007 59
[(Figure_6)TD$FIG]
Figure 6: High [15_TD$DIFF]renewable [40_TD$DIFF]generation in [127_TD$DIFF]game 3B [131_TD$DIFF]makes it [132_TD$DIFF]clear that [133_TD$DIFF]aggregate [45_TD$DIFF]emissionswill be [134_TD$DIFF]far [135_TD$DIFF]under the CO2 [136_TD$DIFF]cap, [137_TD$DIFF]explaining the [138_TD$DIFF]low [33_TD$DIFF]carbon [29_TD$DIFF]costs [139_TD$DIFF]factored in to [128_TD$DIFF]day 2 [140_TD$DIFF]bids(top), [115_TD$DIFF]but [20_TD$DIFF]gencos [141_TD$DIFF]realize by [128_TD$DIFF]day 4 that [142_TD$DIFF]one [143_TD$DIFF]genco [144_TD$DIFF]holds [145_TD$DIFF]almost [146_TD$DIFF]all of the [147_TD$DIFF]allowances, [148_TD$DIFF]so[149_TD$DIFF]they [150_TD$DIFF]need to [151_TD$DIFF]start [152_TD$DIFF]building in a [153_TD$DIFF]high [33_TD$DIFF]carbon [32_TD$DIFF]cost [154_TD$DIFF]anyway ( [155_TD$DIFF]bottom) ( [104_TD$DIFF]bid curves [126_TD$DIFF]shown for[129_TD$DIFF]south region) [130_TD$DIFF].
60
the game in an effort to
accumulate market power in the
REC market, with the hope that
other gencos would refrain from
buying additional renewables as
self-defeating.
Third, it proved difficult for
teams to collectively restrain
purchases of renewables by all of
the other six gencos, though they
certainly tried. All it took was
one team that insisted on
‘‘irrationally’’ buying additional
wind and solar units even after
significant renewable capacity
was in place. The fundamental
incentives problem was that new
renewable additions could be
quite profitable to an individual
generator, as long as nobody
bought too many so that
electricity and REC prices
stayed high. But without some
effective (and perhaps illegal)
way of restraining new
additions, renewables
overcapacity tended to develop,
hurting all generators.
1040-6190/# 2015 Elsevier Inc. All rights reser
D. REC and carbon allowance
trading
If trading is competitive, REC
and carbon allowance prices
should be a function of whether
the electricity market is on target
to meet its respective compliance
targets for renewables and CO2
emissions. If renewable
generation is falling short of the 20
percent RPS target, this should
increase REC prices and stimulate
further wind and solar additions.
If carbon emissions are looking
like they will come in above
the 30-percent-below-business-
as-usual cap, this should boost the
price of allowances and stimulate
mitigation actions, which could
include the incorporation of a
higher price of carbon into
electricity market bids as well as
the acquisition of more renewable
units.
C onsidering REC markets
first, we see that price levels
indeed broadly tracked the level
ved., http://dx.doi.org/10.1016/j.tej.2015.04.007
of wind and solar penetration
(Table 2). The highest average
REC price was observed in the
game with the lowest renewable
energy penetration (2B), while the
lowest average REC price was
observed in the game with
highest renewable energy
penetration (3B).
And yet the detailed trading
patterns shown in Figure 7 also
highlight important departures
from what one would expect in
the competitive market case. The
RPS target was met in the
aggregate in all four games, but
the REC price did not go all
the way to zero in any of them. In
the waning moments of most
games, there ended up being a
limited number of teams with
RECs remaining to sell. Even if
they could not sell all of their
remaining RECs, they were able
to sell the ones they had at a
substantial mark-up due to the
absence of competing suppliers.
In game 3A, for example, Big Coal
made a healthy trading profit by
holding on to its RECs until just
before the close of trading and
then selling large quantities at
high prices even though the RPS
target had been comfortably met
in the aggregate. (In game 3B,
there was such an oversupply of
RECs and abundance of REC
suppliers that this strategy was
not feasible even near the end
of trading.)
T he degree of market power
observed in the carbon
market proved a strong function
of the initial allowance
distribution. Because allowances
The Electricity Journal
[(Figure_7)TD$FIG]
Figure 7: REC Trading in Games 2A/2B and 3A/3B. Each [156_TD$DIFF]circle is a [157_TD$DIFF]trade, with the x-[158_TD$DIFF]coordinate of the [159_TD$DIFF]center [160_TD$DIFF]indicating [79_TD$DIFF]time of the [157_TD$DIFF]trade, y-[158_TD$DIFF]coordinate [160_TD$DIFF]indicating [34_TD$DIFF]price, and [161_TD$DIFF]areaof the [156_TD$DIFF]circle [162_TD$DIFF]proportional to [157_TD$DIFF]trade [163_TD$DIFF]volume. Circles are [61_TD$DIFF]gray for ‘‘A’’ [127_TD$DIFF]game [164_TD$DIFF]trades and [165_TD$DIFF]black for‘‘B’’ [127_TD$DIFF]game [164_TD$DIFF]trades. The [166_TD$DIFF]trading in [167_TD$DIFF]games 2A/2B [168_TD$DIFF]took [169_TD$DIFF]place in [142_TD$DIFF]one [170_TD$DIFF]class [171_TD$DIFF]period; the [166_TD$DIFF]trading in[167_TD$DIFF]games 3A/3B [168_TD$DIFF]took [169_TD$DIFF]place [78_TD$DIFF]over a [171_TD$DIFF]period of [172_TD$DIFF]about [173_TD$DIFF]two [174_TD$DIFF]weeks. Renewable [40_TD$DIFF]generation [175_TD$DIFF]share as a[176_TD$DIFF]percentage of [72_TD$DIFF]total [40_TD$DIFF]generation is shown in [177_TD$DIFF]parentheses for [101_TD$DIFF]each [73_TD$DIFF]game.
M
in games 2A/2B were evenly
distributed among all seven
gencos at the outset, there were
enough allowance sellers and
buyers to produce competitive
outcomes (Figure 8). In game 2A,
once it became clear after day 2 of
the electricity market that
emissions were well within the
cap, allowance prices went to
very low levels. In game 2B,
ay 2015, Vol. 28, Issue 4 1
emissions narrowly exceeded the
cap after the day 2 market was
run, and the resulting scarcity of
allowances drove prices toward
the price cap of $500/ton of CO2.
Carbon market behavior was
very different in games 3A/3B, in
which 100 percent of allowances
were distributed via a sealed
bid, uniform price auction.
Initial allowance supply after
040-6190/# 2015 Elsevier Inc. All rights reserved.,
the auction ended up being quite
concentrated in comparison
with allowance supply in games
2A/2B or REC supply in any
game. In both 3A and 3B, a single
team ended up holding more than
50 percent of the allowances
after the auction, and that team
(Big Gas in game 3A and Big Coal
in game 3B) also ended up being
the provider of all allowances sold
on the final day of trading. As a
result, the final day carbon price
was set not by competitive supply
and demand—after all, there
were plenty of allowances in
aggregate to meet demand in both
games—but rather by a game of
eleventh-hour ‘‘chicken’’ between
monopoly sellers and a limited
number of prospective buyers. As
the final minutes of trading
approached in game 3A, the
number of buyers dwindled and
Big Gas was forced to accept
lower prices. In game 3B,
monopoly supplier Big Coal held
to a firmer line in refusing to
lower allowance prices below
$300. This resulted in heavy losses
for both Big Coal and prospective
buyers that were unable to cover
their emissions, but in a repeated
game this might still have been a
good strategy for Big Coal to show
that it was willing to stick to its
negotiating position even at
significant cost to itself.
T he more equal distribution
of allowances in games 2A/
2B seemed to lead not only to
more competitive trading but
also to earlier trading. One
explanation is that potential
monopolists believe that they will
http://dx.doi.org/10.1016/j.tej.2015.04.007 61
[(Figure_8)TD$FIG]
Figure 8: Carbon Allowance Trading in Games 2A/2B and 3A/3B. Each [156_TD$DIFF]circle is a [157_TD$DIFF]trade,with the x-[158_TD$DIFF]coordinate of the [159_TD$DIFF]center [160_TD$DIFF]indicating [79_TD$DIFF]time of the [157_TD$DIFF]trade, y-[158_TD$DIFF]coordinate [160_TD$DIFF]indicating[34_TD$DIFF]price, and [161_TD$DIFF]area of the [156_TD$DIFF]circle [162_TD$DIFF]proportional to [157_TD$DIFF]trade [163_TD$DIFF]volume. Circles are [61_TD$DIFF]gray for ‘‘A’’ [127_TD$DIFF]game[164_TD$DIFF]trades and [165_TD$DIFF]black for ‘‘B’’ [127_TD$DIFF]game [164_TD$DIFF]trades. The [166_TD$DIFF]trading in [167_TD$DIFF]games 2A/2B [168_TD$DIFF]took [169_TD$DIFF]place in [142_TD$DIFF]one [170_TD$DIFF]class[171_TD$DIFF]period; the [166_TD$DIFF]trading in [167_TD$DIFF]games 3A/3B [168_TD$DIFF]took [169_TD$DIFF]place [78_TD$DIFF]over a [171_TD$DIFF]period of [172_TD$DIFF]about [173_TD$DIFF]two [174_TD$DIFF]weeks. Total[33_TD$DIFF]carbon [45_TD$DIFF]emissions as a [176_TD$DIFF]percentage of the [136_TD$DIFF]cap are [126_TD$DIFF]shown in [177_TD$DIFF]parentheses for [101_TD$DIFF]each [73_TD$DIFF]game.
62
have most leverage to extract
higher prices as the trading
deadline nears. (On the other
hand, buyers can also band
together to try to squeeze the
monopolist with collective buying
power as the deadline
approaches, as occurred in game
3B.) An alternative explanation
that is worthy of further
investigation is that a more even
distribution of allowances, with
1040-6190/# 2015 Elsevier Inc. All rights reser
multiple likely buyers on both
buying and selling sides,
somehow constitutes a ‘‘nudge’’ to
trading that could increase
liquidity in these kinds of markets.
IV. Conclusions
Due to the ease of renewables
acquisition in our game, the
market penetration of utility-scale
ved., http://dx.doi.org/10.1016/j.tej.2015.04.007
renewables was more similar to
that observed in jurisdictions like
Germany with limited
interconnection bottlenecks
and very favorable financial
incentives than to actual
experience thus far in California.13
The game underscored the way
high renewable energy
penetration can depress wholesale
electricity prices, which is a
common phenomenon in
Germany and other European
countries. This situation presents
several practical problems. First, it
makes dispatchable units less
profitable even as they remain
needed for when renewable
energy is not available. (While the
average dispatchable unit in our
game still broke even after the
introduction of renewables, many
higher-variable-cost peaker units
no longer ran at all and thus could
not cover their fixed costs.)
Second, it makes the renewable
units themselves less able to cover
their high fixed costs.
T he game illustrated a few of
the particular challenges of
using an RPS as a tool for stimu-
lating renewable energy
investment. First, REC prices are
inherently unpredictable due to
their sensitivity to the share of
renewable generation relative to
the RPS target.14 We saw this
phenomenon in the game (at least
before teams purchased massive
amounts of wind and solar in
some games); it is also amply
evident in real markets (DOE,
2013). This price uncertainty can
make it difficult to leverage RECs
as a source of long-term value to
The Electricity Journal
M
finance renewables, for example
as part of PPAs (Cory et al.,
2009).15
Second, RECs are a more potent
prod to investment to the extent
that they exist in an environment
of relative policy stability and are
backed by a credible government
commitment to allow their prices
to appreciate significantly if
renewables targets are not met.
This is the situation in our game.
But in California, RPS targets and
rules have periodically changed,
there has been significant
uncertainty about the treatment
of RECs from out-of-state gener-
ation, and transmission con-
straints and other barriers to
renewables that provide utilities
with a (reasonable) excuse for
arguing that they were incapable
of meeting a given target. The
various obstacles to bringing on
new renewable generation in
California are a boon to
generators that might face lower
average wholesale prices in a
more renewables-heavy
environment.
The REC and carbon markets in
our game illustrated the value of
transparency, banking and
borrowing, holding limits, and
other measures to improve
market competitiveness. Because
our game had a significant
number of renewable energy
providers and retailers who
needed RECs, the market was
usually competitive until near the
end of trading when there were
few buyers and sellers remaining.
One lesson for retailers in our
game was that it was safer to
ay 2015, Vol. 28, Issue 4 1
purchase RECs well in advance of
the compliance deadline rather
than risk a high-stakes
negotiation at the end. This
‘‘endgame problem’’ could be
mitigated by allowing banking
and borrowing of RECs across
periods.
A ttempts by players in the
game to accumulate a
monopoly position in RECs might
be mitigated to a degree by more
public disclosure of trades or the
imposition of position limits.
Rules to ensure that renewables
ownership, and therefore REC
production, is not overly
concentrated could help as well.
The game illustrated how the
dynamic nature of RPS
compliance makes REC prices
too volatile to reliably spur
renewable energy investment.
This may suggest that caps and
floors for REC prices would be
useful. A REC price floor would
keep the RPS relevant even if
exogenous factors (for example, a
relaxation of RPS targets or a
downturn in the economy) make
it easier than expected to meet the
040-6190/# 2015 Elsevier Inc. All rights reserved.,
aggregate target. A REC price
ceiling would preserve some
politically feasible incentive for
renewable energy development
even when the RPS target appears
unattainable. In effect, an RPS
with a price cap and floor would
turn into a sliding-scale feed-in
tariff for renewable energy where
the tariff depends on progress
toward the RPS target.
C arbon markets in our game
showed a serious tendency
toward concentration when 100
percent of allowances were
auctioned. In neither of the two
games with full auctioning did
teams intend to accumulate
massive carbon allowance
positions, but quite divergent
expectations among teams about
carbon prices nonetheless led to
this outcome. The exercise of
unilateral market power in carbon
markets that resulted had the
effect of pushing carbon
prices—and thus electricity
prices—far higher than they
would have been based only on
aggregate supply and demand for
allowances. Echoing our
observations above for REC
markets as well as previous
discussion in Thurber and Wolak
(2013), these market power
problems might have been
reduced if our game had
incorporated disclosure rules for
allowance trades, position
limits, banking and borrowing,
and floor and ceiling prices.16
The game revealed the
potential for complex interactions
between the various policies in
play. For example, electricity
http://dx.doi.org/10.1016/j.tej.2015.04.007 63
64
prices spiked when there was
market power in the carbon
market, but only when the
renewable energy fraction was
low enough that thermal units
still set the electricity price most
of the time. High renewable
energy generation discouraged
retailers from focusing as much
on hedging against high prices,
but this could cause them to be
highly exposed if wind and solar
resources are unavailable.
T he simulation raises a
broader question about the
basic wisdom of simultaneously
implementing an RPS and a
cap-and-trade, as in California. If
the cap-and-trade is less stringent
than the RPS, there seems to be
little justification for having both
policies in place unless there is a
carbon price floor. In our game,
we deliberately set the carbon cap
so that it would require additional
mitigation actions beyond simply
meeting the RPS. This could be
rational if the goal is to subsidize
the renewable energy industry up
to a certain point and then seek
the most cost-effective
greenhouse gas emissions
reductions beyond that. It is
important, however, for
policymakers to consciously
consider (and perhaps simulate)
the possible interactions between
the RPS and the cap-and-trade.
E arly applications of this
simulation game as a tool to
help policymakers think through
possible implications of different
policies have been encouraging.
In the spring of 2014 we used
these market simulation tools
1040-6190/# 2015 Elsevier Inc. All rights reser
with a group of regulators from
West Africa who were exploring
possible designs for a wholesale
market in their country. The
game-based approach proved
extremely effective in helping
these energy officials, who came
in with a wide variety of
professional and academic
backgrounds, collectively grapple
with the advantages and
disadvantages of different market
design choices. We will continue
to build out this important
education and policy outreach
component of the work in parallel
with our efforts to run controlled
experiments with students that
rigorously test the impact of
different market rules.&
References
Borenstein, S., Bushnell, J., Wolak, F.A.,Zaragoza-Watkins, M., 2014. Reportof the Market Simulation Group onthe competitive supply/demandbalance in the California allowancemarket and the potential for marketmanipulation. Energy Institute atHaas Working Paper 251. Availableat https://ei.haas.berkeley.edu/pdf/working_papers/WP251.pdf.
ved., http://dx.doi.org/10.1016/j.tej.2015.04.007
Borenstein, S., 2011. The private andpublic economics of renewableelectricity generation. NBER Work-ing Paper 17695. Available athttp://www.nber.org/papers/w17695.
Borenstein, S., Bushnell, J., 2011. Oper-ating Instructions for the ElectricityStrategy Game Version 1.00. Avail-able from Borenstein and Bushnell.
California ISO, 2013. Fast Facts: Whatthe Duck Curve Tells Us AboutManaging a Green Grid. Availableat http://www.caiso.com/Documents/FlexibleResourcesHelpRenewables_FastFacts.pdf.
Cory, K., Canavan, B., Koenig, R., 2009.Power purchase agreement checklistfor state and local governments. Na-tional Renewable Energy Laborato-ry (NREL) Fact Sheet Series onFinancing Renewable Energy Pro-jects. Available at http://www.nrel.gov/docs/fy10osti/46668.pdf.
Denholm, P., Wan, Y.-H., Hummon,M., Mehos, M., 2013. An analysisof concentrating solar power withthermal energy storage in a Califor-nia 33% renewable scenario. NRELTechnical Report TP-6A20-58186.Available at http://www.nrel.gov/docs/fy13osti/58186.pdf.
DOE, 2013. Renewable Energy Certifi-cates (RECs): REC Prices. U.S. De-partment of Energy, EnergyEfficiency & Renewable EnergyWebsite: The Green Power Network.
Available at http://apps3.eere.energy.gov/greenpower/markets/certificates.shtml?page=5.
Thurber, M.C., Wolak, F.A., 2013. Car-bon in the classroom: lesson’s from asimulation of California’s electricitymarket under a stringent cap-and-trade system. Electr. J. 26 (7) 8–21.
Endnotes:
1. In fact, having a patchwork of rulesis more often the norm than theexception in energy andenvironmental policy. California’scombined use of an RPS and acap-and-trade market is one example,but there are many others. Variouscountries already covered by the EUEmissions Trading Scheme (EU ETS)
The Electricity Journal
M
for greenhouse gas emissions havealso deployed specific incentives forrenewable energy, including feed-intariffs (notably in Germany and Spain)and RPS-type policies (as in the case ofthe Renewables Obligation in theU.K.).
2. Trevor Davis coded the newintegrated game.
3. Each renewable technology has adifferent stochastic pattern of outputand average capacity factor modeledon the actual pattern of output duringthe day and capacity factor of actualrenewable generation units employingthis technology.
4. Games 1A/1B did not have fulltrading of allowances and RECs andtherefore are not discussed here.
5. The games described here wereplayed by students in our Winter 2014course on ‘‘Energy Markets andPolicy’’ in the Stanford UniversityGraduate School of Business.
6. These portfolios were originallycreated by Borenstein and Bushnell(2011) to roughly represent theholdings of California powercompanies after restructuring. SeeThurber and Wolak (2013) for asummary of generation portfoliocharacteristics.
7. The exception was that renewablesacquisition was prohibited duringdesignated periods before eachelectricity market day was run so thatteams would know exactly how manyrenewable units would be on thesystem when they decided how to bid
ay 2015, Vol. 28, Issue 4 1
in the capacity of their dispatchablegeneration units, trade forwardcontracts, and make CPP-R decisionsfor the coming day’s electricitymarket.
8. We set the RPS non-compliancepenalty to be equal to the carbonemissions penalty of $500/ton of CO2
multiplied by the emissions rate of thehighest emitting thermal unit in thegame, which is 0.73 tons CO2 perMWh. (Note that emissions rates inour game tend to be lower than thoseof the actual California generatingunits they are modeled after.) Thisapproach treats wind and solaradditions as a carbon mitigationstrategy and assumes for the purposesof setting the RPS penalty that thegeneration they replace would comefrom the highest-emitting thermalunit.
9. Taking advantage of our securehold on power as course instructors,we deliberately made both the RPSand the carbon cap far more stringentthan would be politically tolerable inthe real world, with the goal of makingany effects of these policies morevisible than they might otherwise havebeen.
10. RECs are not actually created untilthe first wind and solar units run, so theearliest any RECs could be traded wasafter the first day of the electricitymarket. In the period since the gamesdescribed here were run, we haveadded ‘‘shorting’’ functionality into thegame that lets players trade RECs andcarbon allowances they do not have.
040-6190/# 2015 Elsevier Inc. All rights reserved.,
11. Note that gencos were able to seeall of the renewables that had alreadybeen brought online before theyacquired new ones.
12. In future versions of the game wecould explore the use of PPA contractsbetween gencos and retailers as analternative mechanism for bringingnew renewable capacity online.
13. It is possible, of course, that gencosin our class would have lost theirenthusiasm for renewables acquisitionif they had repeated the game anumber of additional times. Theprospect for substantial renewablesearnings through RECs in our gamewas not as certain as profitabilityunder a generous feed-in tariff.
14. It is often difficult to forecastexactly how much renewablegeneration will be brought on line in acertain time frame. Renewablegeneration can also be highly sensitiveto the economy (because of its effect oncongestion) and, to some extent,weather. Total electricity consumptionis also a strong function of theeconomy.
15. In theory, derivative instrumentslike fixed-price forward contracts forRECs could help hedge price risk, butthe underlying governmentcommitment to stable RPS policy isprobably too weak to allow a marketfor such derivatives to develop.
16. In future investigations we hope torun a large number of games with andwithout various of these rules in orderto test their effectiveness.
http://dx.doi.org/10.1016/j.tej.2015.04.007 65